Agent superpowers — systematic debugging, test-driven development, planning, and verification. Four skills that transform how the agent approaches complex tasks.

Skills

executing-plans

Execute a multi-step plan generated by writing-plans. Handles failures by retrying the current step, escalating on repeated failure, and reporting completion.

systematic-debugging

Structured approach: isolate failure mode → form hypothesis → instrument → verify → fix. Prevents thrashing on complex bugs by guiding the agent through a disciplined diagnostic loop.

test-driven-development

TDD workflow: write a failing test first, then implement the minimum code to pass it. Keeps scope tight and ensures every new feature has test coverage from day one.

verification-before-completion

Before marking a task done, verify: does the output meet the spec? Are edge cases covered? Were there regressions? Runs a structured checklist before declaring completion.

writing-plans

Decomposes a goal into sequenced, testable steps. Used when the agent receives a complex or ambiguous task — produces a plan before any code is written.

Install

In org template (org.yaml)

plugins:
  - superpowers

From URL (community install)

github://Molecule-AI/molecule-ai-plugin-superpowers

Runtimes

claude_code — primary
deepagents — supported
hermes — supported

Architecture

skills/
  executing-plans/
  systematic-debugging/
  test-driven-development/
  verification-before-completion/
  writing-plans/
adapters/
  claude_code.py      # Registers all skills

Known issues

See known-issues.md.

README.md

superpowers